Upgrade to Pro — share decks privately, control downloads, hide ads and more …

[LASCOT2012] Measure and Manage Flow

6189cdb415d0b7cdbfac8ba2054b2fc1?s=47 Zsolt Fabok
September 21, 2012

[LASCOT2012] Measure and Manage Flow

The fifth iteration of this talk, presented at Lean Agile Scotland 2012.

Please note that the content of this talk changes after each event, because I'm continuously improving it using the feedback and the needs of the audience.

6189cdb415d0b7cdbfac8ba2054b2fc1?s=128

Zsolt Fabok

September 21, 2012
Tweet

Transcript

  1. Measure and Manage Flow in Practice @ZsoltFabok http://zsoltfabok.com/ #lascot12 http://www.leanagilescotland.com/

    by Zsolt Fabok 2012.09.21
  2. None
  3. None
  4. None
  5. #1 Too many open items

  6. WIP point of observation Visualize the situation with Cumulative Flow

    Diagram solved
  7. The Cumulative Flow Diagram Done Started Queued lead time cycle

    time WIP backlog time number of work items It offers more than just the WIP...
  8. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

    ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Q D 3 4 1 2 The simplest way of collecting data:
  9. CFD doesn’t say too much about the throughput A quick

    detour:
  10. work item WIP lead time /* detour */ time

  11. lead time work item throughput 2 WIP /* detour */

  12. work item 2 2 /* detour */

  13. Let’s add more people to the project so that “things

    speed up”! /* detour */
  14. work item 2 2 3 coordination + communication cost /*

    detour */
  15. Actually, “things slow down”, so it was not a good

    idea (solve the right problem instead - systems thinking). End of the detour.
  16. #2 It takes too much time to deliver

  17. lead time CFD was not much help here... ...because we

    didn’t know much about the nature of the lead time
  18. Distribution of lead times days count 0 3 5 8

    10 13 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 22 28 33 56 average median* *Calculation of medians is a popular technique in summary statistics and summarizing statistical data, since it is simple to understand and easy to calculate, while also giving a measure that is more robust in the presence of outlier values
  19. time spent on implementation (hours) time spent waiting (hours) %

    1 63 98 7 57 90 2 62 97 2 62 97 3 61 96 Some examples of work items with 8-day lead time
  20. None
  21. 95% of the lead time was spent on waiting

  22. Distribution of lead times days count 0 3 5 8

    10 13 15 1 4 7 10 13 16 33 Before average 0 3 5 8 10 13 15 1 3 5 7 9 11 13 22 count days After median
  23. #3 Still too many open work items

  24. How many times the item has been rejected

  25. 0 4 8 11 15 31-32 33-34 35-36 37-38 Number

    of rejected work items count week
  26. 0 4 8 11 15 31-32 33-34 35-36 37-38 39-40

    42-43 44-45 Number of rejected work items count week
  27. #4 Being predictable

  28. Sales: “I want to know when the new features can

    hit the market!” Management: “I want to know how much it will cost me!”
  29. All the work items we had so far (~20 work

    items) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ v v v v v v v v
  30. Categorizing them into three groups ~ ~ ~ ~ ~

    ~ ~ ~ ~ ~ ~ ~ v v v v v v v v S M L
  31. The lead time distribution ~ ~ ~ ~ ~ ~

    ~ ~ v v v v M 0 1 1 2 3 3 4 1 2 3 4 5 6 7 8 9 12 13 16 SLA days count Expired
  32. The spent time distribution ~ ~ ~ ~ ~ ~

    ~ ~ v v v v M 0 1 2 3 4 5 6 6 7 8 9 10 SLA hours count
  33. None
  34. #5 Forced improvement

  35. #1 We decided that we would force ourselves to keep

    the SLA #2 Nothing changed. Still the same ratio
  36. Final thoughts

  37. “If you start measuring something you start optimizing it, and

    I know it's the wrong thing to optimize.” A quote which is worth considering http://paulgraham.com/swan.html
  38. The key ideas 1. We develop software not models (value)

    2. Demand first, supply second 3. Observe the system (lead time, throughput) 4. Start measuring, look back if necessary 5. Manage 6. Mind that data expires 7. Goto step 3.
  39. Thank you very much for your attention! http://zsoltfabok.com/ @ZsoltFabok